A simple neural network implemented completely from scratch using NumPy to solve the classic spiral dataset classification problem.
This project manually builds the core components of a neural network — without using frameworks like TensorFlow or PyTorch — to help you understand how neural networks work under the hood.
- Dense (Fully Connected) layers
- ReLU activation function
- Softmax output for multi-class probability
- Categorical Cross-Entropy loss calculation
- Forward propagation implemented with NumPy
This project uses:
- Python
- NumPy
- nnfs (for generating the spiral dataset)
Clone the repository:
git clone https://github.com/yashdeep7733/Spiral-classification-neural-network-3.13.3.git
cd Spiral-classification-neural-network-3.13.3
pip install -r requirements.txt
python spiral-neural-network.py
